Doing some data cleaning and munging. Some of the adhesion limit samples have 6 replications - I can only assume this means I thought the results of the first 3 trials were not correct.
After doing some brief EDA this does seem likely; the ones that have extra reps are for the 20% sand mixes and when using reps 4-6 instead of 1-3 the results are much more similar to one another, which you might expect.
Therefore, I made a new csv file that has reps 1-3 for all samples
that only had 3 reps, and uses reps 4-6 for those which have extras. The
script that generated this “new” csv file is titled
./src/R/data-cleaning/mancino-porcelain-AL-cleaning-INTERACTIVE.R and
the output csv it generates is called
‘ecmdata/derived-data/mancino-intermediate-data/all-porcelain-AL-samples_2021-06-04.csv’.
The data cleaning script that combines all the Atterberg limits data
from the porcelain mixes will therefore need to read in this file from
the derived data directory and some raw data files from the raw data
directory. This is the data cleaning file that should be tracked by
Make.
Today started a rough calculation for the expected water content
differences that would be expected if the change in LL or PL were due
simply to increased surface area. See the file
./computing-expected-water-content-differences-if-due-to-surface-area-alone.R
Labeled experiment 4 as sand sizes vs 3 clay types
Computed specific surface area. Ended up not really using the drawings or the concept of the encapsulating sphere…but here are the paper notes I made on that just in case:
Wrote all day. Made excellent progress. Also worked on table for porosity in Experiment 1.
Wrote for almost 2 hr w/ McNitt. He suggested thinking about splitting this into 2 papers, a Part I and a Part II.
I am going to use git to make a new branch and try this because it will be easy to revert back to where I am now
Today forming new samples for re-doing some of the compression testing. I just don’t feel the data quality is very good, and I am going to be more deliberate about how I am drying the samples down and also about using petrolatum on the plastic disks during the test.
I am going to just continue with unique test IDs, starting at 420.
The first mix I am using today is 30% promound.